IMPACT OF GRAIN BOUNDARIES ON THIN-FILM
PHOTOVOLTAICS
James R. Sites
Physics Department, Colorado State University, Fort Collins, USA
Key words: grain boundaries, GIGS, GdTe Solar cells, rekombinations
Abstract: Grain boundaries in GIGS and GdTe appear to have radically different impact on the solar-cell voltage and fill factor. The proposed explanation is that the valence-band energy in the vicinity of GIGS grain boundaries is shifted downwards, hence keeping the holes in the p-material away from the grain boundaries and potential recombination. Such grain boundaries appearto be an anomalous situation, which would not apply to GdTe and other thin-film polycrystalline materials. The suggestion to avoid grain-boundary recombination in GdTe cells is to construct a fully depleted absorber with an electron reflector layer at its rear surface.
Vpliv mej med zrni na tankoplastno fotovoltaiko
Kjučne besede: meje zrn, rekombinacije, GIGS, GdTe sončne celice
izvleček: Meje kristalnih zrn v polikristalnih GIGS ali GdTe sončnih celicah izkazujejo popolnoma različen vpliv na napetost odprtih sponk in polnilni faktor. Članek utemeljuje razlike z razlago, da je energija valenčnega pasu v bližini mej zrn GIGSa znižana, kar zadržuje vrzeli v p-materialu proč od mej zrn in potencialnih rekombinacij. Takšna sprememba energije valenčnega pasu izgleda neobičajna in se ne pojavlja v GdTe in ostalih polikristalnih tankoplastnih materialih. Da bi se izognili rekombinacijam na mejah zrn v GdTe je potrebno zgraditi popolnoma osiromašeno absorpcijsko plast z elektronsko odbojno plastjo za proste elektrone na zadnji strani.
1 Introduction
Thln-film polycrystalline solar cells have a typical grain, or crystallite, size the order of 1 |im. This means that approximately 1% of their atoms are located on a grain surface, and the likely possibility of extraneous electron states suggests that photovoltaic performance may be seriously compromised. In practice, however, the record-performance Cu(ln,Ga)Se2 (GIGS) solar cells have shown voltages very close to their crystalline counterparts, diode quality factors suggestive of minimal junction recombination, and efficiencies of near 20% /1 /. In contrast, the other leading thin-film material, GdTe, has a record voltage well below that of similar band-gap crystalline cells, a diode-quality factor dominated by recombination, and an efficiency of 16.5%/2/.
If one looks at SEM cross-sections of GIGS and GdTe cells, their micron-scale structure appears remarkably similar. However, instead of the record GdTe efficiency being 3% larger than GIGS, as the two band gaps would predict, it is more than 3% smaller This difference is dramatically illustrated in Fig. 1, where the voltage deficit for the GdTe cell is 200 mV greater than that for GIGS and the primary cause of the difference in efficiency.
The purpose of this presentation is to explore what one might expect for thin-film polycrystalline solar cells, to explain why there is such a dramatic difference between results to date with GIGS and GdTe, and suggest a strategy for eliminating that difference.
2 CIGS: the anomoly
The large voltages achieved by GIGS solar cells suggest that the grain boundaries (GB) have only a minimal effect on their voltage and fill factor More quantitatively, one can calculate the upper limit of grain-boundary recombination velocity (Sgb) that would allow the voltage and fill-factor and fill-factor achieved in Ref. /1/. That value would be the order of 10^ cm/s /3/, which would imply that only about 0.01% of the carriers entering the diode depletion region contribute to the cell's forward current through recombination. That fraction is implausibly low, especially since the comparable Sgb for the highest efficiency GdTe cells, and by inference other micron-scale-grain materials, must be the order of 10® cm/s. A possible explanation is that the carriers (holes) are blocked by an internal potential from reaching the GIGS grain boundaries.
One mechanism that might keep the holes from reaching the GIGS GBs is a positive sheet of charge at the GB. This mechanism is appealing, because the internal surfaces in p-type material may be positively charged. However, this mechanism in itself does not explain the high the voltages observed. In fact, a GB with positive charge will allow additional forward-current recombination and result in smaller voltage than for the equivalent neutral GB /3/. Physically, the same potential that assists collection by channeling the photogenerated electrons and holes will in forward bias provide channels for electrons and holes to flow in the opposite direction, allow greater recombination, increase the forward current, and hence reduce Voc.
A more likely explanation for the benign nature of GIGS GBs is reduced valence-band energy in the vicinity of a
0.2	0.4	0.6
voltage [V]
CN £	
o	
	-10 -
	
'w	
c: 0)	-20 -
-a	
c	
(D	
ä_ 1— .	■30 -
D	
O	
230 mV
CdTe(16.5%).
"GaAs"(E = 1.47 eV)
0.0 0.2 0.4 0.6 0.6 voltage [V]
1,0 1.2
Fig. 1: Current-voltage of record-efficiency CIGS and CdTe celis compared to crystaliine ceils with small adjustments made for band-gap differences.
GB without an equivalent reduction in the conduction band /3/. This scenario is consistent with theoretical analysis of GIGS surfaces and interfaces, where a copper deficit results in an expanded band gap with a lower valence band /4/. It is also supported by experimental measurements of copper deficiency near GBs in CIGS /5/.
Reduction in the valence band by AEv near a GB and the calculated impact on the J-V curves are shown in Fig. 2. For Sgb of both lO'' and 10® cm/s, the voltage and efficiency should approach the values for the GB-free case when AEv is approximately 0.3 eV or larger.
The Fig. 4 curves are reasonably independent of the AEv shape or width, and they are little changed if a modest GB charge is superimposed on the downward expansion of the band gap. Although high voltages and efficiencies experimentally observed with GIGS solar cells could be explained by a very small value of grain-boundary recombination velocity, it seems much more likely that it is the result of the band-gap expansion.
			
			'^Evl
0.6 -
>
0.5 -
s,., = 10"
high i|
-38 -36
-34 I o -32
-30
0.4 o;o AE^ [eV]
Fig. 2: Current-voltage parameters as a function of valence-band reduction (shown schematically above). The values of Sgb are those that would apply if the GB were not shielded by AEv. (From Ref. /3/).
3 CdTe: What can be done?
If thin-film polycrystalline CdTe is the normal situation and has a high GB recombination velocity, is there anything to be done to improve on the voltage and efficiency that has been achieved to date? One strategy would be to substantially reduce the GB defects, increasing both hole density and carrier lifetime (hence reducing Sgb), and essentially make the thin-film CdTe similar to crystalline GaAs. This strategy, however would require nearly a three-order-of-magnitude increase in lifetime and two in carrier density, essentially replicating CIGS values. Such increases are probably not realistic.
An alternative possibility is to expand the conduction band at the back of the CdTe layer, perhaps by alloying with Zn or Mg, to introduce an electron reflector that keeps the photogenerated electrons away from the back metallic contact. This situation is illustrated in Fig. 3, where the barrier has allowed the cell to be in larger forward bias before the forward current limits Voc.
For the electron-reflector strategy to be successful, it is necessary for the CdTe to be fully depleted at the operating voltage so that the internal field will compensate for the relatively low carrier lifetime. In this respect, the low carrier density of CdTe cells (-10^"^ cm®) is advantageous, since the depletion width at zero bias is 1 to2)j,m. To assure the full depletion, however, it would be advisable to keep the CdTe-layer thickness close to 1 |im.
J. R. Sites:
Impact of Grain Boundaries on Thin-film Photovoltaics
(|)g: electron reflector barrier height
i
O) 1— cu
C lu
-1
-2
CdS
0.0 0.5 1.0 1.5 Position (urn)
^bias ^oc I
2.0
Fig. 3: Schematic of electron-reflector strategy to enhance CdTe voltage and efficiency.
Fig. 4 shows the calculated value of voltage for CdTe cells with and without the electron barrier (pe- Without it, the voltage for cells with lifetime x ~ 1 ns is similar to that of today's high-efficiency cells. With the barrier, the voltage should approach that of crystalline material, but further improvement is not predicted for barriers much above 2 eV.
>
1.1
° 0.9 -
0.£
0.7
		(|)e = 0.2 eV
		
		.....^.....—-A—
		
~ 1 1 iiii|-1 1 1 IIIIII-1 1 1 1 niii-1 1 1 iMiii-1 1 1 iiiMi		
0.01
0.1
10
100

Fig. 4: Calculated CdTe voltage as a function of
electron lifetime t and reflection-barrier height (Pe- From Ref. /6/. Full depletion is assumed.
Although the electron-reflector structure would predict a high voltage, its fill-factor will remain lower than optimal unless the CdTe lifetime is increased somewhat above 1 ns. Nevertheless, it should be possible without the need for a major improvement in the quality of thin-film CdTe to reach one volt and 20% efficiency.
4 Conclusions
The reason for the large difference in voltage deficit, and hence efficiency, between polycrystalline CIGS and CdTe solar cells Is that CIGS appears to have a fortuitous reduction in valence band near its grain boundaries that reduces the effective grain-boundary recombination velocity to 1000 cm/s or below. Polycrystalline CdTe does not have this feature, and the CdTe hole density is typically two orders of magnitude below that of CIGS. To significantly increase CdTe voltage, a different device structure based on an electron reflector at the back surface is advised.
Acknowledgments
This work was supported by the U.S. National Renewable Energy Laboratory through its Thin-Film Partnership program. The author is particularly grateful to Dr. Markus Gl-oeckler, Dr. Jun Pan, and Mr. Kuo-Jui (Ray) Hsiao for their contributions to the work presented here.
References
/1/ I. Repins, M.A. Contreras, B. Egaas, C. DeHart, J. Scharf, C.L. Perkins, B. To, and R. Noufi, "19.97o-EfficientZnO/CdS/Culn-GaSes Solar Cell with 81.2% Fill Factor," Prog. Photovolt: Res. Appl. 16, 235(2008).
/2/ X. Wu, J.C. Keane, R.G. Dhere, C. DeHart, D.S. Albin, A. Duda, T.A. Gessert, S. Asher, D.H. Levi, and P. Sheldon, "16.5% Efficient CdS/CdTe Polycrystalline Thin-Film Solar Cell" Proceedings of the 17'" European Photovoltaic Solar Energy Conference, (WIP, Munich, 2001) p. 995
/3/ M. Gloeokler, J.R. Sites, and W.K. Metzger, "Grain-Boundary Recombination in Cu(ln,Ga)Se2 SolarCells," J. Appl. Phys. 98, 113704,(2005).
/4/ C. Perrson and A. Zunger, "Anomalous Grain Boundary Physics in Polycrystalline CulnSe2: The Existence of a Hole Barrier," Phys. Rev Lett. 91, 266401 (2003).
/5/ M.J. Hetzer, Y.M. Strzhemechny, M. Gao, M. Contreras, A. Zunger, and L.J. Brillson, "Direct Observation of Copper Depletion and Potential Changes at Copper Indium Diselenide Grain Boundaries,"/\pp/. Phys. Lett 86, 162105(2005).
/6/ J.R. Sites and J. Pan, "Strategies to Increase CdTe Solar-Cell Voltage," Thin Solid Films 515, 6099-6102 (2007).
James R. Sites
Physics Department, Colorado State University, Fort
Collins, CO 80523 USA